1. Field of the Invention
The present invention relates to a computer program product, system, and method to transfer track format information for tracks in cache at a first processor node to a second process node to which the first processor node is failing over.
2. Description of the Related Art
In a storage environment, a host system may communicate a read/write request to a connected storage system over network channel through a network adaptor. If the data is in a cache of the storage system, i.e., a read hit, then the data may be returned quickly to the host system. This reduces the delay in returning requested data to a host I/O request. However, if the requested data is not in the cache of the storage system, then there may be significant latency realized while the storage system needs to retrieve the requested data from storage to return. Further, the thread or task executing the host read request may have to be context switched and deactivated in order to allow the host system to process further I/O requests. When the data is returned to the read request, then the task must be reactivated and data for the task must be returned to registers and processor cache to allow processing of the returned data for the read request.
In a storage system having two processor nodes, ownership of storage areas or volumes may initially be divided between both processor nodes so each of the processor nodes bears a burden of the I/O requests. In certain situations, one of the processor nodes needs to be taken offline. In such case, a failover may occur from the processor node being taken offline to the surviving processor node to handle I/O requests for those storage areas initially owned by the processor node at which failover is occurring. The failover operation may involve destaging all modified tracks from the processor node failing over and then reassigning ownership of the volumes or storage areas from the failing processor node to the surviving processor node. After the failed processor node becomes available, a failback may occur to reassign the ownership of those storage areas or volumes moved to the surviving processor node back to the other processor node from which the volumes were reassigned to return to the state where both processor nodes are operating and share the storage areas to which I/O requests are directed.
There is a need in the art for improved techniques for processing host read/write requests to the cache in a surviving processor node after a failover.